Path: csiph.com!optima2.xanadu-bbs.net!xanadu-bbs.net!news.glorb.com!usenet.stanford.edu!not-for-mail
From: Dan Douglas <d@ndougl.as>
Newsgroups: gnu.bash.bug
Subject: Re: `foo=1 declare -r foo' prints an error message but its exit status is zero
Date: Fri, 11 Sep 2015 02:04:07 -0500
Lines: 94
Approved: bug-bash@gnu.org
Message-ID: <mailman.901.1441955106.19560.bug-bash@gnu.org>
References: <D447E11E-FFFF-4AE2-B737-1885D6C4FE75@qq.com>
NNTP-Posting-Host: lists.gnu.org
Mime-Version: 1.0
Content-Type: multipart/signed; boundary="nextPart1631624.KXbFBnMrbe"; micalg="pgp-sha1"; protocol="application/pgp-signature"
Cc: ziyunfei <446240525@qq.com>
To: bug-bash@gnu.org
Envelope-to: bug-bash@gnu.org
User-Agent: KMail/5.0.43 pre (Linux/4.2.0; KDE/5.14.0; x86_64; ; )
In-Reply-To: <D447E11E-FFFF-4AE2-B737-1885D6C4FE75@qq.com>
Precedence: list
Xref: csiph.com gnu.bash.bug:11486

--nextPart1631624.KXbFBnMrbe
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="UTF-8"

On Wednesday, September 9, 2015 2:17:30 PM CDT ziyunfei wrote:
> $ foo=1 declare -r foo
> bash: foo: readonly variable
> $ echo $?
> 0
> $ echo $foo
> 1
> 
> Is this a bug? 
> 
> $ bar=1 readonly bar; # throw an error in Bash 4.2, fixed in Bash 4.3

It's a bit more illustrative when you use `readonly' instead of `declare'. In
the former case bash only prints the error in non-POSIX mode because it
modifies the special builtin to unset its value when the command finishes.
`declare' will print an error in either mode. If there's any bug here it's that
bash fails to bail out and executes the echo command anyway.

 $ ( for sh in bash sh ksh mksh zsh dash bb posh; do printf '%-5s x=%5s status=%s\n' "${sh}:" "$("$sh" -c "\${ZSH_VERSION+false}||emulate sh; trap 'echo \${x+unset}' EXIT; $(</dev/fd/3)")" "$?"; done ) 3<<\EOF
x=1 readonly x=2
EOF

bash: x=unset status=0
sh: x: readonly variable
sh:   x=unset status=0
ksh:  x=unset status=0
mksh: x=unset status=0
zsh:  x=unset status=0
dash: x=unset status=0
bb:   x=unset status=0
posh: x=unset status=0

It's possible to trigger a similar error in almost any shell by forcing
readonly to modify the value after the attribute is set.

 $ ( for sh in bash sh ksh mksh zsh dash bb posh; do printf '%-5s x=%5s status=%s\n' "${sh}:" "$("$sh" -c "\${ZSH_VERSION+false}||emulate sh; trap 'echo \${x+unset}' EXIT; $(</dev/fd/3)")" "$?"; done ) 3<<\EOF
readonly x=1 "x=2"
EOF

bash: x: readonly variable
bash: x=unset status=1
sh: x: readonly variable
sh:   x=unset status=1
ksh: readonly: x: is read only
ksh:  x=unset status=1
mksh: read-only: x
mksh: x=unset status=2
zsh:1: read-only variable: x
zsh:  x=      status=1
dash: 1: readonly: x: is read only
dash: x=unset status=2
bb: readonly: line 1: x: is read only
bb:   x=unset status=2
posh: x: is read only
posh: x=      status=1

Interestingly none of my shells trigger this when the same thing is forced by
`command readonly' even if the variable is ultimately modified or unset.

 $ ( for sh in bash sh ksh mksh zsh dash bb posh; do printf '%-5s x=%5s status=%s\n' "${sh}:" "$("$sh" -c "\${ZSH_VERSION+false}||emulate sh; trap 'echo \${x+unset}' EXIT; $(</dev/fd/3)")" "$?"; done ) 3<<\EOF
x=1 command readonly x=2
EOF

bash: x=unset status=0
sh:   x=unset status=0
ksh:  x=unset status=0
mksh: x=      status=0
zsh:  x=      status=0
dash: x=      status=0
bb:   x=unset status=0
posh: x=      status=0

-- 
Dan Douglas
--nextPart1631624.KXbFBnMrbe
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: This is a digitally signed message part.
Content-Transfer-Encoding: 7Bit

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iEYEABECAAYFAlXyfOcACgkQMmyDamdg+Mw19QCff2pLWjwYOFEFuXEvzPakElhN
lZoAoIiVp6SUWOTGF67fGuql0Hxt6O59
=pgEk
-----END PGP SIGNATURE-----

--nextPart1631624.KXbFBnMrbe--